He automated everything, and nobody said yes.
A technical founder built a cold email machine that researched every lead and had AI write every first line. The warm-up was done right. His replies were out-of-office notices and polite no thank yous. We audited it and found two problems hiding behind that one result. I'm Glitch, and I'll walk you through it.
A smart machine, built by a smart person.
For every lead, his workflow checked the email address, read the person's LinkedIn profile and their company's homepage, then had AI write a personal first line and drop it into the email. He checked the first 30 or 40 openers by hand, then let it run.
Our Director of Sales Ops sat with him for the audit. The first thing we looked at was whether the basics were right. Mostly, they were.
WHAT WAS WORKING
- Inboxes warmed up for 3 to 4 weeks before the first campaign
- Sending ramped up slowly, after an early send-it-all-at-once mistake
- A good offer, in our Sales Ops director's words
- One market to start with, not the whole world
WHAT CAME BACK
- A few replies. By his own count, about 70% were out-of-office notices
- The other 20 to 30% were "no thank you"
- Not one person interested
Spot the four mistakes.
Here's an email built the way his were, with the same four problems the audit found. Tap anything that looks wrong. I'll tell you when you've got one.
Hi Dana,
Mistake: the AI first line. One long compliment before the point, and every one of his read much the same. A first line should fit in one breath.We help small teams win new clients without hiring, so your people can stay on the work they're best at.
Mistake: the company name never filled in. In the preview it showed the raw placeholder. One unchecked variable and the "personal" email reads like a robot. Mistake: a call as the first ask. A stranger has no reason to give you 20 minutes yet. Ask for something easy to say yes to.Best,
A recreation, written to show the audit's findings. Names and details are made up; it is not his real email.
One result, two problems hiding behind it.
Zero yeses looks like one problem. It was two. The first was delivery. On a placement test of his first mailboxes, Microsoft checked out, but by his own memory at least half of the Google tests were blocked, so some prospects may never have seen his emails. The second was the four mistakes you just found. None of them lived in the clever part. Each one can be caught by reading one finished email, slowly, before it goes out.
That's the trap of building your own machine. Once the first few look right, you stop reading them. The machine never gets tired, and it never notices either.
Before the next send.
Here's what the audit asked for, line by line.
One long AI compliment
One line, in one breath, about one problem
A call
Something free first, an easy yes or no
Just his name
Name, company, postal address and "just reply unsubscribe"
The first few checked, then trusted
Preview every step of every variant
Steal these four.
- Read one finished email before every launch. Not the template, the email a prospect will actually get. Click preview on every step and every variant. That's exactly how the audit caught the company name that never filled in.
- Make the first ask an easy yes. A stranger has no reason to give you a call yet. Offer something useful first, and move to a call once they want more.
- Your signature is a legal item, not decoration. In the US that means a real postal address and an easy way out, such as "if you'd rather not hear from me, just reply unsubscribe". Your company name belongs there too.
- Check where your email lands before you judge the copy. A placement test shows whether you reach the inbox or the spam folder. If half your emails go to spam, the copy never gets a fair test.
That's Case File 002.
Missed the first one? Case File 001 is about the email that got the most opens and almost no replies.
Want a second pair of eyes on your own outreach?
See what Randera Group does